AOSA/SCST April 2014 Webinar Use of Control Samples in Seed Testing Bryce Callighan Melissa Phillips Monsanto - Waterman AOSA/SCST April 2014 Webinar

Control samples can be a useful tool in the lab to ensure testing parameters were maintained or performed as expected. We will talk about How to select seed How to determine frequency How to use control charts

Question How many attendees or labs currently use control samples? Yes, we use No, we do not

Scenario You are a hard working member of AOSA/SCST who just evaluated a sample with a 56% germination score. Should the sample be retested?

A B C Scenario – Cont. No control sample used Submitted Sample Test Result = 56% germination Result Reported- Unreliable B Control Sample significantly lower than expected Submitted Sample Test Result = 56% germination Process Failure- Retest C Control Sample performed as expected Submitted Sample Test Result = 56% germination Result Reported- Data Reliable

The Why These samples are a “known” or a reference of what should be. These samples are expected to behave in a predicted way in each test situation. When these samples behave in an unexpected way, it can be used as an indicator that there was variation within the testing process. This can be used to verify that a process was executed as required and that the data from that test is reliable.

The Why cont. When you use control samples that behave in an expected manner, “Real” sample data even if unexpected can be a reliable result. Process control samples can be used as a baseline for any improvements or problem solving measures performed in the lab

A Tale of No Reference - True Story Purchased an expensive & award winning bottle of cider I had never had before. Expected- An amazing beverage. Actual Outcome- Worst. Taste. Ever. Assumptions- Bottle was bad, bad batch, it had sat in my trunk on a hot day. Bought another bottle- Still bad Without a control or a reference I was unable to accept that the cider was supposed to taste that way, and had to do an expensive unnecessary “retest" to confirm initial outcome.

When to use control samples ID critical processes that can be affected by minute changes within test constraints Vigor Tests For example: Accelerated Aging Time and Temp play a huge factor in outcomes In your lab you will need to identify and understand Processes at risk for variation Determine what part of the process is at highest risk of compromising the test. This is to be monitored and the frequency established. Per day, per chamber/germinator, per test, per crop/test condition, ect.

How to select control samples Identify two lots to use per crop/test condition Don’t always choose the highest quality lots. High vigor seed may perform well even with additional stress. High vigor seed that always performs well in normal stress conditions won’t alert you if a test under-stressed the seed Don’t go too low in quality Too low in quality could result in a higher standard deviation which could mask any alarms. Why two? Changes of 1 sample may just be due to the seed lot, where changes that affect both lots can easily be attributed to the testing process. Evaluators, abnormal

Some Terminology for Process Control Charts Mean – The average of your control samples Standard Deviation (Sigma) – How much variation there is from the mean. Two sets of tests can both have an average of 80, but have different Standard Deviations Test 1: 90, Test 2: 70. Mean = 80, StDev = 14.14 Test 1: 82, Test 2: 78. Mean = 80, StDev = 2.83

Some Terminology for Process Control Charts UCL and LCL – Upper and Lower Control Limit. This is 3 x the Standard Deviation. Alarm – When a data point is outside of the UCL or LCL. Also when a data trend violates a set of rules. For rules http://www.sqconline.com/about-control-charts

Tips Replace your control samples annually to avoid a sudden drop off in quality If possible, don’t change control samples during your busy season If you have a lot of species in a lab, identify the most susceptible or high impact species. If you only test a species a handful of times, you probably don’t need a control sample of that species Mean shift

Conclusion Regardless of type of lab that you are in… Private, Regulatory, Independent The key to our success is data integrity, reliability and repeatability for our customers The use of control samples enables us to achieve this.

Standard Questions?